Working Surfacelink · 8 Nov 2026
02
A large self-running loom weaves cloth on its own while a few loose threads are lifted to a small panel of numbered levers, where one figure seen from behind pulls a single lever, the whole scene drawn with wide empty margins on every side.
Rescue Now Tech Blog

Leaving task management to AI agents too, until the only work left was answering their confirmation requests

Shimochi · 6 November 2026

How-to · agents · decisions · Agent ownership

Read the original
Takeaway

A Rescue Now engineer lets a coordinating AI agent run their task list and now answers only the decisions it queues, each set out as decision, facts and options.

Summary

Shimochi, an engineer at Rescue Now, published an account in Japanese of handing task management to a coordinating AI agent. Rescue Now is a Japanese company that supplies crisis-management information for disaster preparedness and business continuity. The agent reads the engineer's GitHub and Notion work, passes ready tasks to worker agents and decides their order. It stops for the engineer only when a decision needs a person, through a panel the engineer built.

Key points
  • The coordinator queues design choices and production changes for approval before work starts, and queues pull requests that are ready to merge.
  • It stops before anything another person would see, such as a pull request sent to reviewers or a review comment on a colleague's code.
  • Each queued item shows the decision on its first line, then the facts needed, then the options, with a recommendation when a worker agent is unsure.
  • The engineer answers by item number and a word, in batches, and the coordinator passes the answers on and records them in Notion.
Implication

A team giving AI agents more of its work can define where an agent must stop, and present each stop as decision, facts and options in that order. This is one engineer's setup, and its effect on the wider team is not reported.

Plays
Engineering

When one AI agent hands tasks to other agents, make it wait for an engineer's approval before anything colleagues will see, such as a pull request sent for review.

Engineering

Show each question an AI agent raises with the decision first, then the facts, then the options, so a person can answer by number.

Source issue

8 November 2026: teams using AI agents are moving human review from reading everything an agent produced to settling the decisions it hands back, with evidence attached.